Greg Rolfe

7 months ago · 1 min. reading time · visibility ~10 ·

chat Contact the author

thumb_up Relevant message Comment

Trouble with statistics

For what feels like the majority of my life I have been bombarded with statistics. A set of numbers intended to inform and direct my attention. From the ever-popular 9 out of 10 doctors to the more clinical 30% of the last 43%. You know of what I speak.

But the problem is clear, just how accurate are these numbers? The more you actually look at statistics the more you come to understand just how flexible they are. The smaller the sample rate the greater the potential error, then again the demographic affects the results significantly if you are pooling. The list goes on and on. So just what value does this concept hold?

I have spent my life referring to these numbers as if they were facts or true metrics upon which I could adjust or base decisions on. When in fact they are apparently nothing but marketing tools. This frustrating realization is tantamount to finding out that the tooth fairy isn't real. Who brought me the ever-coveted coins when I left my tooth under my pillow? Now to find out my numbers are just as hypothetical.

Oh, sad day indeed. But now I find myself grown looking once again for that mythical fairy, that warm blanket of numbers. My grasping hands finding instead marketing ploys and guesses. 50 percent of all marriages fail, True? How about that new reality that more and more couples are choosing to ignore the marriage vows altogether. Was that based on the previous statistic? Was that statistic ever accurate?

The problem with statistics is that regardless of the value of the work behind them they are intended to affect our decisions. Statistics are intended to guide our thinking and motivate our choices. But just how valuable are they really. The more I study them more I find that they are just as true as my good friend the tooth fairy.

Please someone tell me I am wrong. I enjoyed my warm blanket.

Blessings!

Trouble with statistics25% [rd 1]
thumb_up Relevant message Comment
Comments
Ken Boddie

Ken Boddie

7 months ago #18

#14
Again, what’s with the insults, Zack? People have different opinions based on different experiences. The only thing I have to ponder is my disappointment that you have chosen to turn this into a personal attack.

Greg Rolfe

Greg Rolfe

7 months ago #17

Ken Boddie thank you both for your positions! My background agrees with Ken while my frustration agrees with Zacharias. Thank you again for adding context to this discussion.

Greg Rolfe

Greg Rolfe

7 months ago #16

#15
Pascal Derrien Dude!!! Sweet!

Pascal Derrien

Pascal Derrien

7 months ago #15

numbers as a mean rather than a finality but I honestly prefer words I find them warmer ....like a blanket :-)

Zacharias 🐝 Voulgaris

Zacharias 🐝 Voulgaris

7 months ago #14

#13
So, your own experience in a very specific field trumps the meta-analysis of an independent researcher? With such a shallow understanding of generalization, no wonder you advocate the merits of this field! If Stats is so great at adding value, why is everyone shifting towards data-driven approaches to analytics, such as Machine Learning and A.I.? Something to ponder upon.

Ken Boddie

Ken Boddie

7 months ago #13

#12
I can only assume from your comment, Zack, that you do not understand how statistic is used to good effect in civil/geotechnical engineering design. My view is based on decades of reproducible work well documented in design procedures and published texts. As it appears to differ from your viewpoint and doubtless from your experience in a different environment, please do not make the assumption that my opinion would benefit from improved accuracy.

Zacharias 🐝 Voulgaris

Zacharias 🐝 Voulgaris

7 months ago #12

Ken Boddie, perhaps Prof. Ioannides' paper on the failure of reproducible results in scientific papers will help you get a more accurate view of this field and how it has failed science repeatedly. In fact, I'd go so far as to say that science would be better off without Stats, at this point, OR with a new more data-driven framework (maybe Stats 2.0) that takes into account reality more than mathematical models. Cheers

Ken Boddie

Ken Boddie

7 months ago #11

I trust, Greg, that my comments in #9 & 10 below may reinstate your faith in statistics, at least when used earnestly and appropriately in legitimate design. 🤗

Ken Boddie

Ken Boddie

7 months ago #10

Continued from below ... I firmly believe that the abusive use of statistics to ‘prove’ preconceived agendas or to manipulate suspect or inappropriate data is not common within the true engineering and scientific domains. I suspect that such abuse is more common, however, within the more mathematically misadventurous dabblings and embellishments of some notoriously factually rubbery sectors of business, such as sales marketing and the tabloid media.

Ken Boddie

Ken Boddie

7 months ago #9

I believe, Greg, that most of us using statistics, within an engineering or scientific context, recognise when a range of data is statistically determinate or not and when the database is representative of our subject matter or not. Statistical determinations are essential tools in the design of many structures that have a limited life, such as pavements. Used correctly, they can assist greatly in the segmentation of the subject field into different environments or components, while assisting the designer to communicate to the client, relationships such as probability of failure versus factor of safety and conservative versus nonconservative approaches ... continued

John Rylance

John Rylance

7 months ago #8

Oops I preessed the wrong key. What I was intending saying statistics that for arguements sake show a 30/70% split can either prove something is working or isnt depending on interpretation. A definite Ah but scenario.

John Rylance

John Rylance

7 months ago #7

There is a joke that Lawyers begin their answer to a question with "It depends", while Accountants reply "What woild you like it to be?" Statisticians combine the two replying, It depends whether you want to prove or disprove it.. For example most statistics centre around percentages. A survey showed 30% of children c

Zacharias 🐝 Voulgaris

Zacharias 🐝 Voulgaris

7 months ago #6

#5
It's easy to abuse what doesn't click with people. Stats manages that more than any other field, partly because of the way it is taught. It is a highly practical field, yet those who teach it approach it like any other kind of Math, focusing on the formulas and the mathy aspects of it. If someone truly understands Stats, it can be quite useful as a way to understand the data (e.g., through the distribution plots, the statistical tests, etc.). It is being abused, for sure. I wish it weren't like that...

Greg Rolfe

Greg Rolfe

7 months ago #5

Hi Zacharias \ud83d\udc1d Voulgaris How are you doing? I studied Statistics in school, though I must admit I did not truly enjoy it. This is probably why I trusted them for so long. Regardless, I am becoming convinced that even the "sub-filed math" is being abused, though I again hope I am wrong.

Zacharias 🐝 Voulgaris

Zacharias 🐝 Voulgaris

7 months ago #4

Perhaps it's best to discern between the commercial Stats and the sub-field of Math that is Stats. The former is more akin to a scam, IMO, and it's heavily abused by those who use it (not always intentionally). The Stats that is part of Math is useful and an intricate part of (data) science. Still, lately, I find that there is literally nothing in analytics that cannot be done by a data-based approach (e.g., Machine Learning), making Stats a nice-to-have, but not an essential tool. Maybe the part of it that deals with the overview of a dataset has some merit, but even that is but a crude generalization, rather than something you can rely on. Having said all that, I enjoy Stats (as a field) and find many of its methods intriguing, even if they are largely obsolete. Cheers

Greg Rolfe

Greg Rolfe

7 months ago #3

#2
Thank you for the link Harvey Lloyd. Yes, indeed some of the Old Testament laws would put quite a kink in branding.

Harvey Lloyd

Harvey Lloyd

7 months ago #2

I too wish for the warm blanket of reasonable truth in data. I awoke one day to realise that all data is selling something. A sad day. IMHO "Branding" has become the gasoline that fuels the flame of "rounding to the ridiculas". Did you know that a Boeing 727 costs 11 cents per passenger mile? What a abuse of statistical analysis. But in branding it is a data point that can sell. I chuckled when i thought of the "Prophets" of old. If they predicted and it didn't come to pass, they were put to death. That would place a kink in the branding of today. http://1.bp.blogspot.com/-2RPpZZu2a-o/UAyaE6xk6XI/AAAAAAAAAPc/uccgjmZKbGI/s1600/Tytler+cycle+cdr+modified+071712.jpg This helped me put all of the "misinformation" in perspective. Data has a purpose and this cycle helped me understand that the purpose of data has changed.

Mohammed Abdul Jawad

Mohammed Abdul Jawad

7 months ago #1

Oftentimes, it’s like forecasts based on factual information, and then predictions and probabilities laced with crispy conjectures and commentaries that create scenarios taking into considerations some calculated assumptions and tactical summaries. And lo! Isn’t there sorcery in statistics that alter human perceptions and lead systems to somewhere?

More articles from Greg Rolfe

View blog
3 weeks ago · 1 min. reading time
Greg Rolfe

Friendship: an art or a skill

I have had the definition of friendship defined as ...

1 month ago · 1 min. reading time
Greg Rolfe

Camping

Camping, a traditional family adventure designed t ...

1 month ago · 1 min. reading time
Greg Rolfe

Returning to a proper focus

There are times when it becomes necessary to just ...